<![CDATA[Stroke is one of the leading causes of death and long-term disability worldwide, making accurate prediction methods essential to support early detection and clinical decision-making. The problem addressed in this study is the lack of evidence regarding which classification algorithm provides the best performance for predicting stroke using the Healthcare Stroke Dataset. This study aims to compare the performance of the Naive Bayes, Random Forest, and Decision Tree algorithms using Orange Data Mining to identify the most effective predictive model. A quantitative approach with a comparative experimental design was employed. The dataset used in this research was the Healthcare Stroke Dataset obtained from Kaggle, consisting of 5,110 records with 12 attributes. The research process included data preprocessing using the Impute widget, feature selection using the Rank widget, classification model development, and model evaluation through 10-fold cross-validation. Performance was assessed using Accuracy, Area Under the Curve (AUC), Precision, Recall, F1-Score, Matthews Correlation Coefficient (MCC), Confusion Matrix, and Receiver Operating Characteristic (ROC) analysis. The results indicate that the Decision Tree algorithm achieved the highest accuracy of 95.1%, followed by Random Forest with 94.8%, while Naive Bayes achieved 92.4%. However, Naive Bayes obtained the highest AUC value of 0.804, demonstrating superior class discrimination capability on an imbalanced dataset. These findings suggest that algorithm selection should not rely solely on accuracy but also consider the model's ability to distinguish between classes consistently. This study contributes to providing recommendations for selecting appropriate classification algorithms to support the development of machine learning-based early stroke prediction systems]]>
